We investigated the possibility of hotspot detection after lithography simulation by using Neural Networks (NN). We
applied the image recognition technique by the NN for hotspot detection and confirmed the possibility by its recognition
rate of the device pattern defects after NN learning.
Various test patterns were prepared for NN learning and we investigated the convergence and the learning time of the
NN. The compositions of the input and the hidden-layers of the NN do not have so much influence on the convergence
of NN, but the initial parameter values of weight setting have predominant effect on the convergence of the NN. There
are correlations among the learning time of the NN, the number of input samples and the number of hidden-layers, so a
certain consideration is required for NN design.
The hotspot recognition rate ranged from 90% to 42%, depending pattern type and learning sample number. Increasing
learning sample number improves the recognition rate. But learning all type patterns leads to 55% recognition, so
learning single type pattern leads to better recognition rate.
This paper reports a technique of reticle inspection incorporating the use of an image filter. In this technique, optical intensity distribution is calculated by optical simulation of electron beam lithography (EB) data or an image file obtained from a SEM photograph to evaluate the printability of defects on a reticle. When an image file is compared with the EB data, the image file has differences at the rounded corners as well as at the areas with defects because the image file is obtained from the reticle pattern. To reduce the differences, an image filter (or reticle filter), which simulates the pattern creation process on a reticle, was applied to the EB data. The simulated EB data is defined as the non-defect reference pattern. The optical intensity and critical dimension (CD) were then obtained. Image files of defects were obtained from the SEM photographs of reticle patterns having various sizes of defects. By applying optical simulation to patterns obtained from the image files, the optical intensity and CD were calculated and compared with those of the reference pattern, and the differences are evaluated. The evaluation results showed that optical intensity and CD changes fluctuate regardless of the size or type of defect. Correlation was confirmed between the differences in optical intensity and the CD changes in the defect area. It was thus concluded that defect printability can be evaluated by the differences in optical intensity obtained from image files.
We conducted an experiment to determine if the use of image filter method for simulation that calculates the distribution of light intensity on a wafer can reduce processing time in comparison to the use of the Fourier transform. The image filter table value is set by changing the value of Gaussian distribution. The image filter method was approximated with the light intensity of optical simulation that keeps accuracy within the range of the allowance. In this experiment, we examined the differences between the distributions calculated using the Fourier transform and the calculation time by varying the sizes of the image filter tables. For the experiment, we used pattern data having a line width that used in the most advanced technology. When the area of pattern data was wide, the experiment revealed that use of the image filter method reduced calculation time by approximately 50 percent or more in comparison to a simulation that used the Fourier transform. As we decreased the size of the image filter tables, the calculation time became shorter, but the differences from the distribution calculated using the Fourier transform became larger. We intend to study the possibility of simulation by expanding the area of pattern data and using the image filter method for simulation-based OPC.
By generating supplementary patterns for EB data and using a system that corrects patten line widths, we improved the shape of a pattern formed on a photomask and the CD linearity. For the EB lithography system, trapezoidal and hammerhead supplementary patterns were applied in order to suppress the increase in EB data volume. As a result, it became possible to reduce the supplementary patterns generated to about 60 percent of the existing serif supplementary patterns. The formed pattern shapes were also equivalent. Since the laser lithography system requires bigger correction pattern shapes than the EB lithography system, triangle supplementary patterns were used. As a result, the corner round was improved with the number of patterns equivalent to that of existing rectangle supplementary patterns. For the CD-linearity, the CD correction amount was set for each line width from the experiment result. For 5 micrometers to 0.7 micrometers patterns on a photomask, a CD-linearity could be achieved within 40nm. We developed the system with above method, when the system is applied to 0.18 micrometers logic contact holes, the elapse time is 1.4 hours and the EB data file size is for 2.5 to 10.8 times the number of original patterns. We judged that it was in the practical level.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.